include/xen/interface/io/blkif.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 /******************************************************************************
   3  * blkif.h
   4  *
   5  * Unified block-device I/O interface for Xen guest OSes.
   6  *
   7  * Copyright (c) 2003-2004, Keir Fraser
   8  */
   9
  10 #ifndef __XEN_PUBLIC_IO_BLKIF_H__
  11 #define __XEN_PUBLIC_IO_BLKIF_H__
  12
  13 #include <xen/interface/io/ring.h>
  14 #include <xen/interface/grant_table.h>
  15
  16 /*
  17  * Front->back notifications: When enqueuing a new request, sending a
  18  * notification can be made conditional on req_event (i.e., the generic
  19  * hold-off mechanism provided by the ring macros). Backends must set
  20  * req_event appropriately (e.g., using RING_FINAL_CHECK_FOR_REQUESTS()).
  21  *
  22  * Back->front notifications: When enqueuing a new response, sending a
  23  * notification can be made conditional on rsp_event (i.e., the generic
  24  * hold-off mechanism provided by the ring macros). Frontends must set
  25  * rsp_event appropriately (e.g., using RING_FINAL_CHECK_FOR_RESPONSES()).
  26  */
  27
  28 typedef uint16_t blkif_vdev_t;
  29 typedef uint64_t blkif_sector_t;
  30
  31 /*
  32  * Multiple hardware queues/rings:
  33  * If supported, the backend will write the key "multi-queue-max-queues" to
  34  * the directory for that vbd, and set its value to the maximum supported
  35  * number of queues.
  36  * Frontends that are aware of this feature and wish to use it can write the
  37  * key "multi-queue-num-queues" with the number they wish to use, which must be
  38  * greater than zero, and no more than the value reported by the backend in
  39  * "multi-queue-max-queues".
  40  *
  41  * For frontends requesting just one queue, the usual event-channel and
  42  * ring-ref keys are written as before, simplifying the backend processing
  43  * to avoid distinguishing between a frontend that doesn't understand the
  44  * multi-queue feature, and one that does, but requested only one queue.
  45  *
  46  * Frontends requesting two or more queues must not write the toplevel
  47  * event-channel and ring-ref keys, instead writing those keys under sub-keys
  48  * having the name "queue-N" where N is the integer ID of the queue/ring for
  49  * which those keys belong. Queues are indexed from zero.
  50  * For example, a frontend with two queues must write the following set of
  51  * queue-related keys:
  52  *
  53  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
  54  * /local/domain/1/device/vbd/0/queue-0 = ""
  55  * /local/domain/1/device/vbd/0/queue-0/ring-ref = "<ring-ref#0>"
  56  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
  57  * /local/domain/1/device/vbd/0/queue-1 = ""
  58  * /local/domain/1/device/vbd/0/queue-1/ring-ref = "<ring-ref#1>"
  59  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
  60  *
  61  * It is also possible to use multiple queues/rings together with
  62  * feature multi-page ring buffer.
  63  * For example, a frontend requests two queues/rings and the size of each ring
  64  * buffer is two pages must write the following set of related keys:
  65  *
  66  * /local/domain/1/device/vbd/0/multi-queue-num-queues = "2"
  67  * /local/domain/1/device/vbd/0/ring-page-order = "1"
  68  * /local/domain/1/device/vbd/0/queue-0 = ""
  69  * /local/domain/1/device/vbd/0/queue-0/ring-ref0 = "<ring-ref#0>"
  70  * /local/domain/1/device/vbd/0/queue-0/ring-ref1 = "<ring-ref#1>"
  71  * /local/domain/1/device/vbd/0/queue-0/event-channel = "<evtchn#0>"
  72  * /local/domain/1/device/vbd/0/queue-1 = ""
  73  * /local/domain/1/device/vbd/0/queue-1/ring-ref0 = "<ring-ref#2>"
  74  * /local/domain/1/device/vbd/0/queue-1/ring-ref1 = "<ring-ref#3>"
  75  * /local/domain/1/device/vbd/0/queue-1/event-channel = "<evtchn#1>"
  76  *
  77  */
  78
  79 /*
  80  * REQUEST CODES.
  81  */
  82 #define BLKIF_OP_READ              0
  83 #define BLKIF_OP_WRITE             1
  84 /*
  85  * Recognised only if "feature-barrier" is present in backend xenbus info.
  86  * The "feature_barrier" node contains a boolean indicating whether barrier
  87  * requests are likely to succeed or fail. Either way, a barrier request
  88  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
  89  * the underlying block-device hardware. The boolean simply indicates whether
  90  * or not it is worthwhile for the frontend to attempt barrier requests.
  91  * If a backend does not recognise BLKIF_OP_WRITE_BARRIER, it should *not*
  92  * create the "feature-barrier" node!
  93  */
  94 #define BLKIF_OP_WRITE_BARRIER     2
  95
  96 /*
  97  * Recognised if "feature-flush-cache" is present in backend xenbus
  98  * info.  A flush will ask the underlying storage hardware to flush its
  99  * non-volatile caches as appropriate.  The "feature-flush-cache" node
 100  * contains a boolean indicating whether flush requests are likely to
 101  * succeed or fail. Either way, a flush request may fail at any time
 102  * with BLKIF_RSP_EOPNOTSUPP if it is unsupported by the underlying
 103  * block-device hardware. The boolean simply indicates whether or not it
 104  * is worthwhile for the frontend to attempt flushes.  If a backend does
 105  * not recognise BLKIF_OP_WRITE_FLUSH_CACHE, it should *not* create the
 106  * "feature-flush-cache" node!
 107  */
 108 #define BLKIF_OP_FLUSH_DISKCACHE   3
 109
 110 /*
 111  * Recognised only if "feature-discard" is present in backend xenbus info.
 112  * The "feature-discard" node contains a boolean indicating whether trim
 113  * (ATA) or unmap (SCSI) - conviently called discard requests are likely
 114  * to succeed or fail. Either way, a discard request
 115  * may fail at any time with BLKIF_RSP_EOPNOTSUPP if it is unsupported by
 116  * the underlying block-device hardware. The boolean simply indicates whether
 117  * or not it is worthwhile for the frontend to attempt discard requests.
 118  * If a backend does not recognise BLKIF_OP_DISCARD, it should *not*
 119  * create the "feature-discard" node!
 120  *
 121  * Discard operation is a request for the underlying block device to mark
 122  * extents to be erased. However, discard does not guarantee that the blocks
 123  * will be erased from the device - it is just a hint to the device
 124  * controller that these blocks are no longer in use. What the device
 125  * controller does with that information is left to the controller.
 126  * Discard operations are passed with sector_number as the
 127  * sector index to begin discard operations at and nr_sectors as the number of
 128  * sectors to be discarded. The specified sectors should be discarded if the
 129  * underlying block device supports trim (ATA) or unmap (SCSI) operations,
 130  * or a BLKIF_RSP_EOPNOTSUPP  should be returned.
 131  * More information about trim/unmap operations at:
 132  * http://t13.org/Documents/UploadedDocuments/docs2008/
 133  *     e07154r6-Data_Set_Management_Proposal_for_ATA-ACS2.doc
 134  * http://www.seagate.com/staticfiles/support/disc/manuals/
 135  *     Interface%20manuals/100293068c.pdf
 136  * The backend can optionally provide three extra XenBus attributes to
 137  * further optimize the discard functionality:
 138  * 'discard-alignment' - Devices that support discard functionality may
 139  * internally allocate space in units that are bigger than the exported
 140  * logical block size. The discard-alignment parameter indicates how many bytes
 141  * the beginning of the partition is offset from the internal allocation unit's
 142  * natural alignment.
 143  * 'discard-granularity'  - Devices that support discard functionality may
 144  * internally allocate space using units that are bigger than the logical block
 145  * size. The discard-granularity parameter indicates the size of the internal
 146  * allocation unit in bytes if reported by the device. Otherwise the
 147  * discard-granularity will be set to match the device's physical block size.
 148  * 'discard-secure' - All copies of the discarded sectors (potentially created
 149  * by garbage collection) must also be erased.  To use this feature, the flag
 150  * BLKIF_DISCARD_SECURE must be set in the blkif_request_trim.
 151  */
 152 #define BLKIF_OP_DISCARD           5
 153
 154 /*
 155  * Recognized if "feature-max-indirect-segments" in present in the backend
 156  * xenbus info. The "feature-max-indirect-segments" node contains the maximum
 157  * number of segments allowed by the backend per request. If the node is
 158  * present, the frontend might use blkif_request_indirect structs in order to
 159  * issue requests with more than BLKIF_MAX_SEGMENTS_PER_REQUEST (11). The
 160  * maximum number of indirect segments is fixed by the backend, but the
 161  * frontend can issue requests with any number of indirect segments as long as
 162  * it's less than the number provided by the backend. The indirect_grefs field
 163  * in blkif_request_indirect should be filled by the frontend with the
 164  * grant references of the pages that are holding the indirect segments.
 165  * These pages are filled with an array of blkif_request_segment that hold the
 166  * information about the segments. The number of indirect pages to use is
 167  * determined by the number of segments an indirect request contains. Every
 168  * indirect page can contain a maximum of
 169  * (PAGE_SIZE / sizeof(struct blkif_request_segment)) segments, so to
 170  * calculate the number of indirect pages to use we have to do
 171  * ceil(indirect_segments / (PAGE_SIZE / sizeof(struct blkif_request_segment))).
 172  *
 173  * If a backend does not recognize BLKIF_OP_INDIRECT, it should *not*
 174  * create the "feature-max-indirect-segments" node!
 175  */
 176 #define BLKIF_OP_INDIRECT          6
 177
 178 /*
 179  * Maximum scatter/gather segments per request.
 180  * This is carefully chosen so that sizeof(struct blkif_ring) <= PAGE_SIZE.
 181  * NB. This could be 12 if the ring indexes weren't stored in the same page.
 182  */
 183 #define BLKIF_MAX_SEGMENTS_PER_REQUEST 11
 184
 185 #define BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST 8
 186
 187 struct blkif_request_segment {
 188                 grant_ref_t gref;        /* reference to I/O buffer frame        */
 189                 /* @first_sect: first sector in frame to transfer (inclusive).   */
 190                 /* @last_sect: last sector in frame to transfer (inclusive).     */
 191                 uint8_t     first_sect, last_sect;
 192 };
 193
 194 struct blkif_request_rw {
 195         uint8_t        nr_segments;  /* number of segments                   */
 196         blkif_vdev_t   handle;       /* only for read/write requests         */
 197 #ifndef CONFIG_X86_32
 198         uint32_t       _pad1;        /* offsetof(blkif_request,u.rw.id) == 8 */
 199 #endif
 200         uint64_t       id;           /* private guest value, echoed in resp  */
 201         blkif_sector_t sector_number;/* start sector idx on disk (r/w only)  */
 202         struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
 203 } __attribute__((__packed__));
 204
 205 struct blkif_request_discard {
 206         uint8_t        flag;         /* BLKIF_DISCARD_SECURE or zero.        */
 207 #define BLKIF_DISCARD_SECURE (1<<0)  /* ignored if discard-secure=0          */
 208         blkif_vdev_t   _pad1;        /* only for read/write requests         */
 209 #ifndef CONFIG_X86_32
 210         uint32_t       _pad2;        /* offsetof(blkif_req..,u.discard.id)==8*/
 211 #endif
 212         uint64_t       id;           /* private guest value, echoed in resp  */
 213         blkif_sector_t sector_number;
 214         uint64_t       nr_sectors;
 215         uint8_t        _pad3;
 216 } __attribute__((__packed__));
 217
 218 struct blkif_request_other {
 219         uint8_t      _pad1;
 220         blkif_vdev_t _pad2;        /* only for read/write requests         */
 221 #ifndef CONFIG_X86_32
 222         uint32_t     _pad3;        /* offsetof(blkif_req..,u.other.id)==8*/
 223 #endif
 224         uint64_t     id;           /* private guest value, echoed in resp  */
 225 } __attribute__((__packed__));
 226
 227 struct blkif_request_indirect {
 228         uint8_t        indirect_op;
 229         uint16_t       nr_segments;
 230 #ifndef CONFIG_X86_32
 231         uint32_t       _pad1;        /* offsetof(blkif_...,u.indirect.id) == 8 */
 232 #endif
 233         uint64_t       id;
 234         blkif_sector_t sector_number;
 235         blkif_vdev_t   handle;
 236         uint16_t       _pad2;
 237         grant_ref_t    indirect_grefs[BLKIF_MAX_INDIRECT_PAGES_PER_REQUEST];
 238 #ifndef CONFIG_X86_32
 239         uint32_t      _pad3;         /* make it 64 byte aligned */
 240 #else
 241         uint64_t      _pad3;         /* make it 64 byte aligned */
 242 #endif
 243 } __attribute__((__packed__));
 244
 245 struct blkif_request {
 246         uint8_t        operation;    /* BLKIF_OP_???                         */
 247         union {
 248                 struct blkif_request_rw rw;
 249                 struct blkif_request_discard discard;
 250                 struct blkif_request_other other;
 251                 struct blkif_request_indirect indirect;
 252         } u;
 253 } __attribute__((__packed__));
 254
 255 struct blkif_response {
 256         uint64_t        id;              /* copied from request */
 257         uint8_t         operation;       /* copied from request */
 258         int16_t         status;          /* BLKIF_RSP_???       */
 259 };
 260
 261 /*
 262  * STATUS RETURN CODES.
 263  */
 264  /* Operation not supported (only happens on barrier writes). */
 265 #define BLKIF_RSP_EOPNOTSUPP  -2
 266  /* Operation failed for some unspecified reason (-EIO). */
 267 #define BLKIF_RSP_ERROR       -1
 268  /* Operation completed successfully. */
 269 #define BLKIF_RSP_OKAY         0
 270
 271 /*
 272  * Generate blkif ring structures and types.
 273  */
 274
 275 DEFINE_RING_TYPES(blkif, struct blkif_request, struct blkif_response);
 276
 277 #define VDISK_CDROM        0x1
 278 #define VDISK_REMOVABLE    0x2
 279 #define VDISK_READONLY     0x4
 280
 281 /* Xen-defined major numbers for virtual disks, they look strangely
 282  * familiar */
 283 #define XEN_IDE0_MAJOR  3
 284 #define XEN_IDE1_MAJOR  22
 285 #define XEN_SCSI_DISK0_MAJOR    8
 286 #define XEN_SCSI_DISK1_MAJOR    65
 287 #define XEN_SCSI_DISK2_MAJOR    66
 288 #define XEN_SCSI_DISK3_MAJOR    67
 289 #define XEN_SCSI_DISK4_MAJOR    68
 290 #define XEN_SCSI_DISK5_MAJOR    69
 291 #define XEN_SCSI_DISK6_MAJOR    70
 292 #define XEN_SCSI_DISK7_MAJOR    71
 293 #define XEN_SCSI_DISK8_MAJOR    128
 294 #define XEN_SCSI_DISK9_MAJOR    129
 295 #define XEN_SCSI_DISK10_MAJOR   130
 296 #define XEN_SCSI_DISK11_MAJOR   131
 297 #define XEN_SCSI_DISK12_MAJOR   132
 298 #define XEN_SCSI_DISK13_MAJOR   133
 299 #define XEN_SCSI_DISK14_MAJOR   134
 300 #define XEN_SCSI_DISK15_MAJOR   135
 301
 302 #endif /* __XEN_PUBLIC_IO_BLKIF_H__ */